Constructing a High Performance and Low Power Campus Research Cyberinfrastructure at IU

Project ID
FG-545
Project Categories
Computer Science
Abstract
Recently Indiana University was cited for its leadership in cyberinfrastructure and data as part of NSF's 6 Big Ideas. This project will enhance this leadership in both research in the software tools and hardware architecture as well as the use of cyberinfrastructure. The campus cyberinfrastructure such as supercomputer supported by University Information Technology Services (UITS) will be used largely in simulation, data analytics and engineering arena, as well as mission-critical initiatives such as the Precision Health Initiative of IU Grand Challenges. The Intelligent Systems Engineering Department (ISE) will be a significant user of nanotechnology and bioengineering while researching cyberinfrastructure technology in computer architecture and intelligent systems on data analytics and HPC. We expect substantial interest throughout the University including physics, chemistry, bioinformatics, network science and security, due to the important positioning of computing technologies to modernize applications. What is not known is what a high performance, low power and cost-effective IU cyberinfrastructure from both software and hardware aspects actually take place when these computations are performed. Accordingly, there is an urgent need to identify and evaluate the applications underlie these computations. Without this, our empirical understanding of campus cyberinfrastructure will continue to have a blind spot at the technical capabilities and scale for the dominant computation workloads needed and predict the most important computations to occur. In this project, we propose to a holistic methodology to improve the software and hardware foundations for a high performance and low power campus research cyberinfrastructure at IU. In this project, we propose to a holistic methodology to improve the software and hardware foundations for a high performance and low power campus research cyberinfrastructure at IU. (Task1:) The work creates the first research benchmark suite covering diverse IU needs for simulation and data ana lytics. (Task2:) Furthermore, the proposal also identifies the pros and cons of emerging hardware platforms running the IU suite. (Task3:) At last, this work is the first to construct a software framework for the IU suite to fully utilize state-of-the-art emerging hardware platforms. Preliminary results suggest the feasibility and promise of our approaches to benchmarking, characterization, and optimization. Together, our research contributions and preliminary results comprise a viable proposal that considerably advances the state of the art.
Use of FutureSystems
We will run some benchmark applications on the FutureSystems.
Scale of Use
We need 24 hours CPU time per week.